~ Enhancing AI Learning Data Infrastructure, Driving Social Implementation Toward 100 Mass-Produced Units and 10,000 Cumulative Hours of Field Operation ~

ZEALS took the stage on Monday, August 31, 2026, at the final presentation event for the "Physical AI Development Support Program by AWS Japan" hosted by Amazon Web Services Japan G.K. (hereinafter "AWS Japan"), where it demonstrated and exhibited the semi-domestic compact humanoid robot "D1."

At the presentation, ZEALS introduced the design philosophy of D1 tailored to Japan's field environments, integrated development achievements across hardware, software, and AI models, and the AI learning data infrastructure built with technical support from AWS.

Leveraging these results, ZEALS will promote the field deployment of D1 and continuous improvement of AI models based on operational data. Aiming for 100 mass-produced units and 10,000 cumulative hours of field operation in FY2026, we will advance the social implementation and business expansion of Physical AI.

1. About the "Physical AI Development Support Program by AWS Japan"

This program is an initiative conducted by AWS Japan to support the development and social implementation of Physical AI in Japan.

Over approximately six months starting from March 2026, the 51 selected companies received expert technical support, cost optimization assistance, community building, and Go-To-Market support.

At the final presentation event held at the AWS Azabudai Hills office on August 31, 14 companies across 13 projects presented their development results and initiatives toward social implementation.

Through participation in this program, ZEALS worked on building data processing infrastructure and enhancing learning environments required for developing humanoid AI models, and presented these achievements.

2. Realizing Physical AI Evolving from Japanese Field Environments

ZEALS believes that workplace environments in Japan—facing acute labor shortages in medical/nursing care, manufacturing/logistics, and service industries—serve as a crucial foundation to accelerate the social implementation and evolution of Physical AI.

Advancing Physical AI requires real-world data obtained through robots operating in actual environments. It is essential to capture varying conditions across workplaces—such as human movement patterns, equipment layouts, item shapes, and operational force control—and utilize them to train and improve AI models.

Meanwhile, indoor environments in Japanese healthcare and nursing care facilities present specific requirements for robot deployment, including limited hallway widths, proximity to humans, and safety considerations.

The semi-domestic compact humanoid "D1" was developed to operate effectively in these environments and continuously collect operational data from the field.

D1 features a compact wheel-based mobile base approximately 48 cm wide combined with a dual-arm structure, designed with an emphasis on safety and practical utility in Japanese indoor environments.

Beyond simply delivering robots to operational sites, ZEALS aims to establish a development cycle where data collected from field operations is used to train and improve AI models, redeploying further sophisticated robots back to the field.

3. Enhancing AI Learning Data Infrastructure with AWS Technical Support

At the start of the program, ZEALS focused mainly on software development using third-party robots. Subsequently, we developed the semi-domestic compact humanoid "D1," the integrated software suite "Omakase OS" supporting conversation and autonomous navigation, and the manipulation intelligence platform "Omakase Zen," establishing an integrated development structure for hardware, software, and AI models.

In developing Omakase Zen specifically, ZEALS built a data processing infrastructure with technical support from AWS to process field-collected data for AI training and evaluation. By optimizing distributed data processing and conversion pipelines, data conversion speed was improved by 10x compared to previous methods.

As D1 deployments expand, the volume of data collected by robots is expected to grow significantly. Thus, establishing an infrastructure capable of efficiently processing collected data for AI model training and evaluation is a critical element supporting future business scaling.

By leveraging this newly built data processing pipeline, we will accelerate the development cycle of continuously feeding field data into learning and evaluation, redeploying improved models back to D1.

Through this approach, we aim to simultaneously expand D1 deployments and drive continuous AI model evolution, progressively enhancing adaptability across diverse tasks and environments.

4. Future Outlook | Expanding D1 Business Through 100 Mass-Produced Units and 10,000 Hours of Field Operation

ZEALS has set targets for FY2026 of mass-producing 100 D1 units and achieving a cumulative 10,000 hours of operation in actual operational environments.

What matters most in this goal is not just the number of units sold, but how extensively the robots operate in actual working environments to accumulate operational field data.

Field operational data—capturing human and object movements, layout variations, and unexpected scenarios difficult to fully replicate in R&D settings—serves as a vital asset for increasing AI model autonomy and task capability.

From August 26 to 28, 2026, ZEALS deployed two D1 units to Juko Osu Hospital in Nagoya City, completing a total of 35 hours of field operation over three days. The trial verified adaptability across five hospital tasks: visitor guidance, lightweight item transport, tea serving assistance, tray clearing assistance, and exercise leading.

Moving forward, we will further advance trials in the medical and nursing care sectors while expanding into various domains facing labor shortages, including logistics, manufacturing, and service industries.

Furthermore, using the data infrastructure established with AWS technical support, we will comprehensively drive the end-to-end cycle from field operations and data collection to AI model training, refinement, and redeployment.

**"As deployment advances, more data accumulates, AI models evolve, enabling handling of an even broader range of tasks."** By building this continuous feedback loop, ZEALS aims to enhance D1's practical utility and business competitiveness.

By combining the domain expertise of Japanese workplaces with Physical AI technology, ZEALS strives to realize a society where humans and robots work side by side.

Related Information

AWS Official | "Physical AI Development Support Program by AWS Japan" Final Presentation Results
https://aws.amazon.com/jp/blogs/news/aws-japan-physical-ai-development-support-program-results/

D1 Proof of Concept at Juko Osu Hospital
https://omakaserobotics.ai/news/jukoosu-poc

D1 Official Website
https://omakaserobotics.ai/